Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Copyright (C) 2001, 2002 Sistina Software (UK) Limited. | |
2b06cfff | 3 | * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved. |
1da177e4 LT |
4 | * |
5 | * This file is released under the GPL. | |
6 | */ | |
7 | ||
8 | #include "dm.h" | |
9 | #include "dm-bio-list.h" | |
51e5b2bd | 10 | #include "dm-uevent.h" |
1da177e4 LT |
11 | |
12 | #include <linux/init.h> | |
13 | #include <linux/module.h> | |
48c9c27b | 14 | #include <linux/mutex.h> |
1da177e4 LT |
15 | #include <linux/moduleparam.h> |
16 | #include <linux/blkpg.h> | |
17 | #include <linux/bio.h> | |
18 | #include <linux/buffer_head.h> | |
19 | #include <linux/mempool.h> | |
20 | #include <linux/slab.h> | |
21 | #include <linux/idr.h> | |
3ac51e74 | 22 | #include <linux/hdreg.h> |
2056a782 | 23 | #include <linux/blktrace_api.h> |
aa129a22 | 24 | #include <linux/smp_lock.h> |
1da177e4 | 25 | |
72d94861 AK |
26 | #define DM_MSG_PREFIX "core" |
27 | ||
1da177e4 LT |
28 | static const char *_name = DM_NAME; |
29 | ||
30 | static unsigned int major = 0; | |
31 | static unsigned int _major = 0; | |
32 | ||
f32c10b0 | 33 | static DEFINE_SPINLOCK(_minor_lock); |
1da177e4 LT |
34 | /* |
35 | * One of these is allocated per bio. | |
36 | */ | |
37 | struct dm_io { | |
38 | struct mapped_device *md; | |
39 | int error; | |
1da177e4 | 40 | atomic_t io_count; |
6ae2fa67 | 41 | struct bio *bio; |
3eaf840e | 42 | unsigned long start_time; |
1da177e4 LT |
43 | }; |
44 | ||
45 | /* | |
46 | * One of these is allocated per target within a bio. Hopefully | |
47 | * this will be simplified out one day. | |
48 | */ | |
028867ac | 49 | struct dm_target_io { |
1da177e4 LT |
50 | struct dm_io *io; |
51 | struct dm_target *ti; | |
52 | union map_info info; | |
53 | }; | |
54 | ||
55 | union map_info *dm_get_mapinfo(struct bio *bio) | |
56 | { | |
17b2f66f | 57 | if (bio && bio->bi_private) |
028867ac | 58 | return &((struct dm_target_io *)bio->bi_private)->info; |
17b2f66f | 59 | return NULL; |
1da177e4 LT |
60 | } |
61 | ||
ba61fdd1 JM |
62 | #define MINOR_ALLOCED ((void *)-1) |
63 | ||
1da177e4 LT |
64 | /* |
65 | * Bits for the md->flags field. | |
66 | */ | |
67 | #define DMF_BLOCK_IO 0 | |
68 | #define DMF_SUSPENDED 1 | |
aa8d7c2f | 69 | #define DMF_FROZEN 2 |
fba9f90e | 70 | #define DMF_FREEING 3 |
5c6bd75d | 71 | #define DMF_DELETING 4 |
2e93ccc1 | 72 | #define DMF_NOFLUSH_SUSPENDING 5 |
1da177e4 | 73 | |
304f3f6a MB |
74 | /* |
75 | * Work processed by per-device workqueue. | |
76 | */ | |
77 | struct dm_wq_req { | |
78 | enum { | |
79 | DM_WQ_FLUSH_ALL, | |
80 | DM_WQ_FLUSH_DEFERRED, | |
81 | } type; | |
82 | struct work_struct work; | |
83 | struct mapped_device *md; | |
84 | void *context; | |
85 | }; | |
86 | ||
1da177e4 | 87 | struct mapped_device { |
2ca3310e | 88 | struct rw_semaphore io_lock; |
e61290a4 | 89 | struct mutex suspend_lock; |
2e93ccc1 | 90 | spinlock_t pushback_lock; |
1da177e4 LT |
91 | rwlock_t map_lock; |
92 | atomic_t holders; | |
5c6bd75d | 93 | atomic_t open_count; |
1da177e4 LT |
94 | |
95 | unsigned long flags; | |
96 | ||
165125e1 | 97 | struct request_queue *queue; |
1da177e4 | 98 | struct gendisk *disk; |
7e51f257 | 99 | char name[16]; |
1da177e4 LT |
100 | |
101 | void *interface_ptr; | |
102 | ||
103 | /* | |
104 | * A list of ios that arrived while we were suspended. | |
105 | */ | |
106 | atomic_t pending; | |
107 | wait_queue_head_t wait; | |
74859364 | 108 | struct bio_list deferred; |
2e93ccc1 | 109 | struct bio_list pushback; |
1da177e4 | 110 | |
304f3f6a MB |
111 | /* |
112 | * Processing queue (flush/barriers) | |
113 | */ | |
114 | struct workqueue_struct *wq; | |
115 | ||
1da177e4 LT |
116 | /* |
117 | * The current mapping. | |
118 | */ | |
119 | struct dm_table *map; | |
120 | ||
121 | /* | |
122 | * io objects are allocated from here. | |
123 | */ | |
124 | mempool_t *io_pool; | |
125 | mempool_t *tio_pool; | |
126 | ||
9faf400f SB |
127 | struct bio_set *bs; |
128 | ||
1da177e4 LT |
129 | /* |
130 | * Event handling. | |
131 | */ | |
132 | atomic_t event_nr; | |
133 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
134 | atomic_t uevent_seq; |
135 | struct list_head uevent_list; | |
136 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
137 | |
138 | /* | |
139 | * freeze/thaw support require holding onto a super block | |
140 | */ | |
141 | struct super_block *frozen_sb; | |
e39e2e95 | 142 | struct block_device *suspended_bdev; |
3ac51e74 DW |
143 | |
144 | /* forced geometry settings */ | |
145 | struct hd_geometry geometry; | |
1da177e4 LT |
146 | }; |
147 | ||
148 | #define MIN_IOS 256 | |
e18b890b CL |
149 | static struct kmem_cache *_io_cache; |
150 | static struct kmem_cache *_tio_cache; | |
1da177e4 | 151 | |
1da177e4 LT |
152 | static int __init local_init(void) |
153 | { | |
154 | int r; | |
155 | ||
1da177e4 | 156 | /* allocate a slab for the dm_ios */ |
028867ac | 157 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 LT |
158 | if (!_io_cache) |
159 | return -ENOMEM; | |
160 | ||
161 | /* allocate a slab for the target ios */ | |
028867ac | 162 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
1da177e4 LT |
163 | if (!_tio_cache) { |
164 | kmem_cache_destroy(_io_cache); | |
165 | return -ENOMEM; | |
166 | } | |
167 | ||
51e5b2bd MA |
168 | r = dm_uevent_init(); |
169 | if (r) { | |
170 | kmem_cache_destroy(_tio_cache); | |
171 | kmem_cache_destroy(_io_cache); | |
172 | return r; | |
173 | } | |
174 | ||
1da177e4 LT |
175 | _major = major; |
176 | r = register_blkdev(_major, _name); | |
177 | if (r < 0) { | |
178 | kmem_cache_destroy(_tio_cache); | |
179 | kmem_cache_destroy(_io_cache); | |
51e5b2bd | 180 | dm_uevent_exit(); |
1da177e4 LT |
181 | return r; |
182 | } | |
183 | ||
184 | if (!_major) | |
185 | _major = r; | |
186 | ||
187 | return 0; | |
188 | } | |
189 | ||
190 | static void local_exit(void) | |
191 | { | |
192 | kmem_cache_destroy(_tio_cache); | |
193 | kmem_cache_destroy(_io_cache); | |
00d59405 | 194 | unregister_blkdev(_major, _name); |
51e5b2bd | 195 | dm_uevent_exit(); |
1da177e4 LT |
196 | |
197 | _major = 0; | |
198 | ||
199 | DMINFO("cleaned up"); | |
200 | } | |
201 | ||
b9249e55 | 202 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
203 | local_init, |
204 | dm_target_init, | |
205 | dm_linear_init, | |
206 | dm_stripe_init, | |
945fa4d2 | 207 | dm_kcopyd_init, |
1da177e4 LT |
208 | dm_interface_init, |
209 | }; | |
210 | ||
b9249e55 | 211 | static void (*_exits[])(void) = { |
1da177e4 LT |
212 | local_exit, |
213 | dm_target_exit, | |
214 | dm_linear_exit, | |
215 | dm_stripe_exit, | |
945fa4d2 | 216 | dm_kcopyd_exit, |
1da177e4 LT |
217 | dm_interface_exit, |
218 | }; | |
219 | ||
220 | static int __init dm_init(void) | |
221 | { | |
222 | const int count = ARRAY_SIZE(_inits); | |
223 | ||
224 | int r, i; | |
225 | ||
226 | for (i = 0; i < count; i++) { | |
227 | r = _inits[i](); | |
228 | if (r) | |
229 | goto bad; | |
230 | } | |
231 | ||
232 | return 0; | |
233 | ||
234 | bad: | |
235 | while (i--) | |
236 | _exits[i](); | |
237 | ||
238 | return r; | |
239 | } | |
240 | ||
241 | static void __exit dm_exit(void) | |
242 | { | |
243 | int i = ARRAY_SIZE(_exits); | |
244 | ||
245 | while (i--) | |
246 | _exits[i](); | |
247 | } | |
248 | ||
249 | /* | |
250 | * Block device functions | |
251 | */ | |
252 | static int dm_blk_open(struct inode *inode, struct file *file) | |
253 | { | |
254 | struct mapped_device *md; | |
255 | ||
fba9f90e JM |
256 | spin_lock(&_minor_lock); |
257 | ||
1da177e4 | 258 | md = inode->i_bdev->bd_disk->private_data; |
fba9f90e JM |
259 | if (!md) |
260 | goto out; | |
261 | ||
5c6bd75d AK |
262 | if (test_bit(DMF_FREEING, &md->flags) || |
263 | test_bit(DMF_DELETING, &md->flags)) { | |
fba9f90e JM |
264 | md = NULL; |
265 | goto out; | |
266 | } | |
267 | ||
1da177e4 | 268 | dm_get(md); |
5c6bd75d | 269 | atomic_inc(&md->open_count); |
fba9f90e JM |
270 | |
271 | out: | |
272 | spin_unlock(&_minor_lock); | |
273 | ||
274 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
275 | } |
276 | ||
277 | static int dm_blk_close(struct inode *inode, struct file *file) | |
278 | { | |
279 | struct mapped_device *md; | |
280 | ||
281 | md = inode->i_bdev->bd_disk->private_data; | |
5c6bd75d | 282 | atomic_dec(&md->open_count); |
1da177e4 LT |
283 | dm_put(md); |
284 | return 0; | |
285 | } | |
286 | ||
5c6bd75d AK |
287 | int dm_open_count(struct mapped_device *md) |
288 | { | |
289 | return atomic_read(&md->open_count); | |
290 | } | |
291 | ||
292 | /* | |
293 | * Guarantees nothing is using the device before it's deleted. | |
294 | */ | |
295 | int dm_lock_for_deletion(struct mapped_device *md) | |
296 | { | |
297 | int r = 0; | |
298 | ||
299 | spin_lock(&_minor_lock); | |
300 | ||
301 | if (dm_open_count(md)) | |
302 | r = -EBUSY; | |
303 | else | |
304 | set_bit(DMF_DELETING, &md->flags); | |
305 | ||
306 | spin_unlock(&_minor_lock); | |
307 | ||
308 | return r; | |
309 | } | |
310 | ||
3ac51e74 DW |
311 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
312 | { | |
313 | struct mapped_device *md = bdev->bd_disk->private_data; | |
314 | ||
315 | return dm_get_geometry(md, geo); | |
316 | } | |
317 | ||
aa129a22 MB |
318 | static int dm_blk_ioctl(struct inode *inode, struct file *file, |
319 | unsigned int cmd, unsigned long arg) | |
320 | { | |
321 | struct mapped_device *md; | |
322 | struct dm_table *map; | |
323 | struct dm_target *tgt; | |
324 | int r = -ENOTTY; | |
325 | ||
326 | /* We don't really need this lock, but we do need 'inode'. */ | |
327 | unlock_kernel(); | |
328 | ||
329 | md = inode->i_bdev->bd_disk->private_data; | |
330 | ||
331 | map = dm_get_table(md); | |
332 | ||
333 | if (!map || !dm_table_get_size(map)) | |
334 | goto out; | |
335 | ||
336 | /* We only support devices that have a single target */ | |
337 | if (dm_table_get_num_targets(map) != 1) | |
338 | goto out; | |
339 | ||
340 | tgt = dm_table_get_target(map, 0); | |
341 | ||
342 | if (dm_suspended(md)) { | |
343 | r = -EAGAIN; | |
344 | goto out; | |
345 | } | |
346 | ||
347 | if (tgt->type->ioctl) | |
348 | r = tgt->type->ioctl(tgt, inode, file, cmd, arg); | |
349 | ||
350 | out: | |
351 | dm_table_put(map); | |
352 | ||
353 | lock_kernel(); | |
354 | return r; | |
355 | } | |
356 | ||
028867ac | 357 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
358 | { |
359 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
360 | } | |
361 | ||
028867ac | 362 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
363 | { |
364 | mempool_free(io, md->io_pool); | |
365 | } | |
366 | ||
028867ac | 367 | static struct dm_target_io *alloc_tio(struct mapped_device *md) |
1da177e4 LT |
368 | { |
369 | return mempool_alloc(md->tio_pool, GFP_NOIO); | |
370 | } | |
371 | ||
028867ac | 372 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
373 | { |
374 | mempool_free(tio, md->tio_pool); | |
375 | } | |
376 | ||
3eaf840e JNN |
377 | static void start_io_acct(struct dm_io *io) |
378 | { | |
379 | struct mapped_device *md = io->md; | |
c9959059 | 380 | int cpu; |
3eaf840e JNN |
381 | |
382 | io->start_time = jiffies; | |
383 | ||
074a7aca TH |
384 | cpu = part_stat_lock(); |
385 | part_round_stats(cpu, &dm_disk(md)->part0); | |
386 | part_stat_unlock(); | |
387 | dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending); | |
3eaf840e JNN |
388 | } |
389 | ||
390 | static int end_io_acct(struct dm_io *io) | |
391 | { | |
392 | struct mapped_device *md = io->md; | |
393 | struct bio *bio = io->bio; | |
394 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 395 | int pending, cpu; |
3eaf840e JNN |
396 | int rw = bio_data_dir(bio); |
397 | ||
074a7aca TH |
398 | cpu = part_stat_lock(); |
399 | part_round_stats(cpu, &dm_disk(md)->part0); | |
400 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
401 | part_stat_unlock(); | |
3eaf840e | 402 | |
074a7aca TH |
403 | dm_disk(md)->part0.in_flight = pending = |
404 | atomic_dec_return(&md->pending); | |
3eaf840e JNN |
405 | |
406 | return !pending; | |
407 | } | |
408 | ||
1da177e4 LT |
409 | /* |
410 | * Add the bio to the list of deferred io. | |
411 | */ | |
412 | static int queue_io(struct mapped_device *md, struct bio *bio) | |
413 | { | |
2ca3310e | 414 | down_write(&md->io_lock); |
1da177e4 LT |
415 | |
416 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 417 | up_write(&md->io_lock); |
1da177e4 LT |
418 | return 1; |
419 | } | |
420 | ||
421 | bio_list_add(&md->deferred, bio); | |
422 | ||
2ca3310e | 423 | up_write(&md->io_lock); |
1da177e4 LT |
424 | return 0; /* deferred successfully */ |
425 | } | |
426 | ||
427 | /* | |
428 | * Everyone (including functions in this file), should use this | |
429 | * function to access the md->map field, and make sure they call | |
430 | * dm_table_put() when finished. | |
431 | */ | |
432 | struct dm_table *dm_get_table(struct mapped_device *md) | |
433 | { | |
434 | struct dm_table *t; | |
435 | ||
436 | read_lock(&md->map_lock); | |
437 | t = md->map; | |
438 | if (t) | |
439 | dm_table_get(t); | |
440 | read_unlock(&md->map_lock); | |
441 | ||
442 | return t; | |
443 | } | |
444 | ||
3ac51e74 DW |
445 | /* |
446 | * Get the geometry associated with a dm device | |
447 | */ | |
448 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
449 | { | |
450 | *geo = md->geometry; | |
451 | ||
452 | return 0; | |
453 | } | |
454 | ||
455 | /* | |
456 | * Set the geometry of a device. | |
457 | */ | |
458 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
459 | { | |
460 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
461 | ||
462 | if (geo->start > sz) { | |
463 | DMWARN("Start sector is beyond the geometry limits."); | |
464 | return -EINVAL; | |
465 | } | |
466 | ||
467 | md->geometry = *geo; | |
468 | ||
469 | return 0; | |
470 | } | |
471 | ||
1da177e4 LT |
472 | /*----------------------------------------------------------------- |
473 | * CRUD START: | |
474 | * A more elegant soln is in the works that uses the queue | |
475 | * merge fn, unfortunately there are a couple of changes to | |
476 | * the block layer that I want to make for this. So in the | |
477 | * interests of getting something for people to use I give | |
478 | * you this clearly demarcated crap. | |
479 | *---------------------------------------------------------------*/ | |
480 | ||
2e93ccc1 KU |
481 | static int __noflush_suspending(struct mapped_device *md) |
482 | { | |
483 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
484 | } | |
485 | ||
1da177e4 LT |
486 | /* |
487 | * Decrements the number of outstanding ios that a bio has been | |
488 | * cloned into, completing the original io if necc. | |
489 | */ | |
858119e1 | 490 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 491 | { |
2e93ccc1 KU |
492 | unsigned long flags; |
493 | ||
494 | /* Push-back supersedes any I/O errors */ | |
495 | if (error && !(io->error > 0 && __noflush_suspending(io->md))) | |
1da177e4 LT |
496 | io->error = error; |
497 | ||
498 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
499 | if (io->error == DM_ENDIO_REQUEUE) { |
500 | /* | |
501 | * Target requested pushing back the I/O. | |
502 | * This must be handled before the sleeper on | |
503 | * suspend queue merges the pushback list. | |
504 | */ | |
505 | spin_lock_irqsave(&io->md->pushback_lock, flags); | |
506 | if (__noflush_suspending(io->md)) | |
507 | bio_list_add(&io->md->pushback, io->bio); | |
508 | else | |
509 | /* noflush suspend was interrupted. */ | |
510 | io->error = -EIO; | |
511 | spin_unlock_irqrestore(&io->md->pushback_lock, flags); | |
512 | } | |
513 | ||
3eaf840e | 514 | if (end_io_acct(io)) |
1da177e4 LT |
515 | /* nudge anyone waiting on suspend queue */ |
516 | wake_up(&io->md->wait); | |
517 | ||
2e93ccc1 KU |
518 | if (io->error != DM_ENDIO_REQUEUE) { |
519 | blk_add_trace_bio(io->md->queue, io->bio, | |
520 | BLK_TA_COMPLETE); | |
521 | ||
6712ecf8 | 522 | bio_endio(io->bio, io->error); |
2e93ccc1 | 523 | } |
2056a782 | 524 | |
1da177e4 LT |
525 | free_io(io->md, io); |
526 | } | |
527 | } | |
528 | ||
6712ecf8 | 529 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
530 | { |
531 | int r = 0; | |
028867ac | 532 | struct dm_target_io *tio = bio->bi_private; |
9faf400f | 533 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
534 | dm_endio_fn endio = tio->ti->type->end_io; |
535 | ||
1da177e4 LT |
536 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
537 | error = -EIO; | |
538 | ||
539 | if (endio) { | |
540 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
541 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
542 | /* | |
543 | * error and requeue request are handled | |
544 | * in dec_pending(). | |
545 | */ | |
1da177e4 | 546 | error = r; |
45cbcd79 KU |
547 | else if (r == DM_ENDIO_INCOMPLETE) |
548 | /* The target will handle the io */ | |
6712ecf8 | 549 | return; |
45cbcd79 KU |
550 | else if (r) { |
551 | DMWARN("unimplemented target endio return value: %d", r); | |
552 | BUG(); | |
553 | } | |
1da177e4 LT |
554 | } |
555 | ||
9faf400f SB |
556 | dec_pending(tio->io, error); |
557 | ||
558 | /* | |
559 | * Store md for cleanup instead of tio which is about to get freed. | |
560 | */ | |
561 | bio->bi_private = md->bs; | |
562 | ||
1da177e4 | 563 | bio_put(bio); |
9faf400f | 564 | free_tio(md, tio); |
1da177e4 LT |
565 | } |
566 | ||
567 | static sector_t max_io_len(struct mapped_device *md, | |
568 | sector_t sector, struct dm_target *ti) | |
569 | { | |
570 | sector_t offset = sector - ti->begin; | |
571 | sector_t len = ti->len - offset; | |
572 | ||
573 | /* | |
574 | * Does the target need to split even further ? | |
575 | */ | |
576 | if (ti->split_io) { | |
577 | sector_t boundary; | |
578 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
579 | - offset; | |
580 | if (len > boundary) | |
581 | len = boundary; | |
582 | } | |
583 | ||
584 | return len; | |
585 | } | |
586 | ||
587 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 588 | struct dm_target_io *tio) |
1da177e4 LT |
589 | { |
590 | int r; | |
2056a782 | 591 | sector_t sector; |
9faf400f | 592 | struct mapped_device *md; |
1da177e4 LT |
593 | |
594 | /* | |
595 | * Sanity checks. | |
596 | */ | |
597 | BUG_ON(!clone->bi_size); | |
598 | ||
599 | clone->bi_end_io = clone_endio; | |
600 | clone->bi_private = tio; | |
601 | ||
602 | /* | |
603 | * Map the clone. If r == 0 we don't need to do | |
604 | * anything, the target has assumed ownership of | |
605 | * this io. | |
606 | */ | |
607 | atomic_inc(&tio->io->io_count); | |
2056a782 | 608 | sector = clone->bi_sector; |
1da177e4 | 609 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 610 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 611 | /* the bio has been remapped so dispatch it */ |
2056a782 | 612 | |
17b2f66f | 613 | blk_add_trace_remap(bdev_get_queue(clone->bi_bdev), clone, |
c7149d6b AB |
614 | tio->io->bio->bi_bdev->bd_dev, |
615 | clone->bi_sector, sector); | |
2056a782 | 616 | |
1da177e4 | 617 | generic_make_request(clone); |
2e93ccc1 KU |
618 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
619 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
620 | md = tio->io->md; |
621 | dec_pending(tio->io, r); | |
622 | /* | |
623 | * Store bio_set for cleanup. | |
624 | */ | |
625 | clone->bi_private = md->bs; | |
1da177e4 | 626 | bio_put(clone); |
9faf400f | 627 | free_tio(md, tio); |
45cbcd79 KU |
628 | } else if (r) { |
629 | DMWARN("unimplemented target map return value: %d", r); | |
630 | BUG(); | |
1da177e4 LT |
631 | } |
632 | } | |
633 | ||
634 | struct clone_info { | |
635 | struct mapped_device *md; | |
636 | struct dm_table *map; | |
637 | struct bio *bio; | |
638 | struct dm_io *io; | |
639 | sector_t sector; | |
640 | sector_t sector_count; | |
641 | unsigned short idx; | |
642 | }; | |
643 | ||
3676347a PO |
644 | static void dm_bio_destructor(struct bio *bio) |
645 | { | |
9faf400f SB |
646 | struct bio_set *bs = bio->bi_private; |
647 | ||
648 | bio_free(bio, bs); | |
3676347a PO |
649 | } |
650 | ||
1da177e4 LT |
651 | /* |
652 | * Creates a little bio that is just does part of a bvec. | |
653 | */ | |
654 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
655 | unsigned short idx, unsigned int offset, | |
9faf400f | 656 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
657 | { |
658 | struct bio *clone; | |
659 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
660 | ||
9faf400f | 661 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 662 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
663 | *clone->bi_io_vec = *bv; |
664 | ||
665 | clone->bi_sector = sector; | |
666 | clone->bi_bdev = bio->bi_bdev; | |
667 | clone->bi_rw = bio->bi_rw; | |
668 | clone->bi_vcnt = 1; | |
669 | clone->bi_size = to_bytes(len); | |
670 | clone->bi_io_vec->bv_offset = offset; | |
671 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 672 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 LT |
673 | |
674 | return clone; | |
675 | } | |
676 | ||
677 | /* | |
678 | * Creates a bio that consists of range of complete bvecs. | |
679 | */ | |
680 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
681 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 682 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
683 | { |
684 | struct bio *clone; | |
685 | ||
9faf400f SB |
686 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
687 | __bio_clone(clone, bio); | |
688 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
689 | clone->bi_sector = sector; |
690 | clone->bi_idx = idx; | |
691 | clone->bi_vcnt = idx + bv_count; | |
692 | clone->bi_size = to_bytes(len); | |
693 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
694 | ||
695 | return clone; | |
696 | } | |
697 | ||
512875bd | 698 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
699 | { |
700 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
701 | struct dm_target *ti; |
702 | sector_t len = 0, max; | |
028867ac | 703 | struct dm_target_io *tio; |
1da177e4 | 704 | |
512875bd JN |
705 | ti = dm_table_find_target(ci->map, ci->sector); |
706 | if (!dm_target_is_valid(ti)) | |
707 | return -EIO; | |
708 | ||
709 | max = max_io_len(ci->md, ci->sector, ti); | |
710 | ||
1da177e4 LT |
711 | /* |
712 | * Allocate a target io object. | |
713 | */ | |
714 | tio = alloc_tio(ci->md); | |
715 | tio->io = ci->io; | |
716 | tio->ti = ti; | |
717 | memset(&tio->info, 0, sizeof(tio->info)); | |
718 | ||
719 | if (ci->sector_count <= max) { | |
720 | /* | |
721 | * Optimise for the simple case where we can do all of | |
722 | * the remaining io with a single clone. | |
723 | */ | |
724 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
725 | bio->bi_vcnt - ci->idx, ci->sector_count, |
726 | ci->md->bs); | |
1da177e4 LT |
727 | __map_bio(ti, clone, tio); |
728 | ci->sector_count = 0; | |
729 | ||
730 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
731 | /* | |
732 | * There are some bvecs that don't span targets. | |
733 | * Do as many of these as possible. | |
734 | */ | |
735 | int i; | |
736 | sector_t remaining = max; | |
737 | sector_t bv_len; | |
738 | ||
739 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
740 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
741 | ||
742 | if (bv_len > remaining) | |
743 | break; | |
744 | ||
745 | remaining -= bv_len; | |
746 | len += bv_len; | |
747 | } | |
748 | ||
9faf400f SB |
749 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
750 | ci->md->bs); | |
1da177e4 LT |
751 | __map_bio(ti, clone, tio); |
752 | ||
753 | ci->sector += len; | |
754 | ci->sector_count -= len; | |
755 | ci->idx = i; | |
756 | ||
757 | } else { | |
758 | /* | |
d2044a94 | 759 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
760 | */ |
761 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
762 | sector_t remaining = to_sector(bv->bv_len); |
763 | unsigned int offset = 0; | |
1da177e4 | 764 | |
d2044a94 AK |
765 | do { |
766 | if (offset) { | |
767 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
768 | if (!dm_target_is_valid(ti)) |
769 | return -EIO; | |
770 | ||
d2044a94 | 771 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 772 | |
d2044a94 AK |
773 | tio = alloc_tio(ci->md); |
774 | tio->io = ci->io; | |
775 | tio->ti = ti; | |
776 | memset(&tio->info, 0, sizeof(tio->info)); | |
777 | } | |
778 | ||
779 | len = min(remaining, max); | |
780 | ||
781 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
782 | bv->bv_offset + offset, len, |
783 | ci->md->bs); | |
d2044a94 AK |
784 | |
785 | __map_bio(ti, clone, tio); | |
786 | ||
787 | ci->sector += len; | |
788 | ci->sector_count -= len; | |
789 | offset += to_bytes(len); | |
790 | } while (remaining -= len); | |
1da177e4 | 791 | |
1da177e4 LT |
792 | ci->idx++; |
793 | } | |
512875bd JN |
794 | |
795 | return 0; | |
1da177e4 LT |
796 | } |
797 | ||
798 | /* | |
799 | * Split the bio into several clones. | |
800 | */ | |
9e4e5f87 | 801 | static int __split_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
802 | { |
803 | struct clone_info ci; | |
512875bd | 804 | int error = 0; |
1da177e4 LT |
805 | |
806 | ci.map = dm_get_table(md); | |
9e4e5f87 MB |
807 | if (unlikely(!ci.map)) |
808 | return -EIO; | |
1da177e4 LT |
809 | |
810 | ci.md = md; | |
811 | ci.bio = bio; | |
812 | ci.io = alloc_io(md); | |
813 | ci.io->error = 0; | |
814 | atomic_set(&ci.io->io_count, 1); | |
815 | ci.io->bio = bio; | |
816 | ci.io->md = md; | |
817 | ci.sector = bio->bi_sector; | |
818 | ci.sector_count = bio_sectors(bio); | |
819 | ci.idx = bio->bi_idx; | |
820 | ||
3eaf840e | 821 | start_io_acct(ci.io); |
512875bd JN |
822 | while (ci.sector_count && !error) |
823 | error = __clone_and_map(&ci); | |
1da177e4 LT |
824 | |
825 | /* drop the extra reference count */ | |
512875bd | 826 | dec_pending(ci.io, error); |
1da177e4 | 827 | dm_table_put(ci.map); |
9e4e5f87 MB |
828 | |
829 | return 0; | |
1da177e4 LT |
830 | } |
831 | /*----------------------------------------------------------------- | |
832 | * CRUD END | |
833 | *---------------------------------------------------------------*/ | |
834 | ||
f6fccb12 MB |
835 | static int dm_merge_bvec(struct request_queue *q, |
836 | struct bvec_merge_data *bvm, | |
837 | struct bio_vec *biovec) | |
838 | { | |
839 | struct mapped_device *md = q->queuedata; | |
840 | struct dm_table *map = dm_get_table(md); | |
841 | struct dm_target *ti; | |
842 | sector_t max_sectors; | |
5037108a | 843 | int max_size = 0; |
f6fccb12 MB |
844 | |
845 | if (unlikely(!map)) | |
5037108a | 846 | goto out; |
f6fccb12 MB |
847 | |
848 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
849 | if (!dm_target_is_valid(ti)) |
850 | goto out_table; | |
f6fccb12 MB |
851 | |
852 | /* | |
853 | * Find maximum amount of I/O that won't need splitting | |
854 | */ | |
855 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
856 | (sector_t) BIO_MAX_SECTORS); | |
857 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
858 | if (max_size < 0) | |
859 | max_size = 0; | |
860 | ||
861 | /* | |
862 | * merge_bvec_fn() returns number of bytes | |
863 | * it can accept at this offset | |
864 | * max is precomputed maximal io size | |
865 | */ | |
866 | if (max_size && ti->type->merge) | |
867 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
868 | ||
b01cd5ac | 869 | out_table: |
5037108a MP |
870 | dm_table_put(map); |
871 | ||
872 | out: | |
f6fccb12 MB |
873 | /* |
874 | * Always allow an entire first page | |
875 | */ | |
876 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
877 | max_size = biovec->bv_len; | |
878 | ||
f6fccb12 MB |
879 | return max_size; |
880 | } | |
881 | ||
1da177e4 LT |
882 | /* |
883 | * The request function that just remaps the bio built up by | |
884 | * dm_merge_bvec. | |
885 | */ | |
165125e1 | 886 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 887 | { |
9e4e5f87 | 888 | int r = -EIO; |
12f03a49 | 889 | int rw = bio_data_dir(bio); |
1da177e4 | 890 | struct mapped_device *md = q->queuedata; |
c9959059 | 891 | int cpu; |
1da177e4 | 892 | |
07a83c47 SB |
893 | /* |
894 | * There is no use in forwarding any barrier request since we can't | |
895 | * guarantee it is (or can be) handled by the targets correctly. | |
896 | */ | |
897 | if (unlikely(bio_barrier(bio))) { | |
6712ecf8 | 898 | bio_endio(bio, -EOPNOTSUPP); |
07a83c47 SB |
899 | return 0; |
900 | } | |
901 | ||
2ca3310e | 902 | down_read(&md->io_lock); |
1da177e4 | 903 | |
074a7aca TH |
904 | cpu = part_stat_lock(); |
905 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
906 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
907 | part_stat_unlock(); | |
12f03a49 | 908 | |
1da177e4 LT |
909 | /* |
910 | * If we're suspended we have to queue | |
911 | * this io for later. | |
912 | */ | |
913 | while (test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 914 | up_read(&md->io_lock); |
1da177e4 | 915 | |
9e4e5f87 MB |
916 | if (bio_rw(bio) != READA) |
917 | r = queue_io(md, bio); | |
1da177e4 | 918 | |
9e4e5f87 MB |
919 | if (r <= 0) |
920 | goto out_req; | |
1da177e4 LT |
921 | |
922 | /* | |
923 | * We're in a while loop, because someone could suspend | |
924 | * before we get to the following read lock. | |
925 | */ | |
2ca3310e | 926 | down_read(&md->io_lock); |
1da177e4 LT |
927 | } |
928 | ||
9e4e5f87 | 929 | r = __split_bio(md, bio); |
2ca3310e | 930 | up_read(&md->io_lock); |
9e4e5f87 MB |
931 | |
932 | out_req: | |
933 | if (r < 0) | |
934 | bio_io_error(bio); | |
935 | ||
1da177e4 LT |
936 | return 0; |
937 | } | |
938 | ||
165125e1 | 939 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
940 | { |
941 | struct mapped_device *md = q->queuedata; | |
942 | struct dm_table *map = dm_get_table(md); | |
943 | ||
944 | if (map) { | |
945 | dm_table_unplug_all(map); | |
946 | dm_table_put(map); | |
947 | } | |
948 | } | |
949 | ||
950 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
951 | { | |
952 | int r; | |
953 | struct mapped_device *md = (struct mapped_device *) congested_data; | |
954 | struct dm_table *map = dm_get_table(md); | |
955 | ||
956 | if (!map || test_bit(DMF_BLOCK_IO, &md->flags)) | |
957 | r = bdi_bits; | |
958 | else | |
959 | r = dm_table_any_congested(map, bdi_bits); | |
960 | ||
961 | dm_table_put(map); | |
962 | return r; | |
963 | } | |
964 | ||
965 | /*----------------------------------------------------------------- | |
966 | * An IDR is used to keep track of allocated minor numbers. | |
967 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
968 | static DEFINE_IDR(_minor_idr); |
969 | ||
2b06cfff | 970 | static void free_minor(int minor) |
1da177e4 | 971 | { |
f32c10b0 | 972 | spin_lock(&_minor_lock); |
1da177e4 | 973 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 974 | spin_unlock(&_minor_lock); |
1da177e4 LT |
975 | } |
976 | ||
977 | /* | |
978 | * See if the device with a specific minor # is free. | |
979 | */ | |
cf13ab8e | 980 | static int specific_minor(int minor) |
1da177e4 LT |
981 | { |
982 | int r, m; | |
983 | ||
984 | if (minor >= (1 << MINORBITS)) | |
985 | return -EINVAL; | |
986 | ||
62f75c2f JM |
987 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
988 | if (!r) | |
989 | return -ENOMEM; | |
990 | ||
f32c10b0 | 991 | spin_lock(&_minor_lock); |
1da177e4 LT |
992 | |
993 | if (idr_find(&_minor_idr, minor)) { | |
994 | r = -EBUSY; | |
995 | goto out; | |
996 | } | |
997 | ||
ba61fdd1 | 998 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 999 | if (r) |
1da177e4 | 1000 | goto out; |
1da177e4 LT |
1001 | |
1002 | if (m != minor) { | |
1003 | idr_remove(&_minor_idr, m); | |
1004 | r = -EBUSY; | |
1005 | goto out; | |
1006 | } | |
1007 | ||
1008 | out: | |
f32c10b0 | 1009 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1010 | return r; |
1011 | } | |
1012 | ||
cf13ab8e | 1013 | static int next_free_minor(int *minor) |
1da177e4 | 1014 | { |
2b06cfff | 1015 | int r, m; |
1da177e4 | 1016 | |
1da177e4 | 1017 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1018 | if (!r) |
1019 | return -ENOMEM; | |
1020 | ||
f32c10b0 | 1021 | spin_lock(&_minor_lock); |
1da177e4 | 1022 | |
ba61fdd1 | 1023 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1024 | if (r) |
1da177e4 | 1025 | goto out; |
1da177e4 LT |
1026 | |
1027 | if (m >= (1 << MINORBITS)) { | |
1028 | idr_remove(&_minor_idr, m); | |
1029 | r = -ENOSPC; | |
1030 | goto out; | |
1031 | } | |
1032 | ||
1033 | *minor = m; | |
1034 | ||
1035 | out: | |
f32c10b0 | 1036 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1037 | return r; |
1038 | } | |
1039 | ||
1040 | static struct block_device_operations dm_blk_dops; | |
1041 | ||
1042 | /* | |
1043 | * Allocate and initialise a blank device with a given minor. | |
1044 | */ | |
2b06cfff | 1045 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1046 | { |
1047 | int r; | |
cf13ab8e | 1048 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1049 | void *old_md; |
1da177e4 LT |
1050 | |
1051 | if (!md) { | |
1052 | DMWARN("unable to allocate device, out of memory."); | |
1053 | return NULL; | |
1054 | } | |
1055 | ||
10da4f79 | 1056 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1057 | goto bad_module_get; |
10da4f79 | 1058 | |
1da177e4 | 1059 | /* get a minor number for the dev */ |
2b06cfff | 1060 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1061 | r = next_free_minor(&minor); |
2b06cfff | 1062 | else |
cf13ab8e | 1063 | r = specific_minor(minor); |
1da177e4 | 1064 | if (r < 0) |
6ed7ade8 | 1065 | goto bad_minor; |
1da177e4 | 1066 | |
2ca3310e | 1067 | init_rwsem(&md->io_lock); |
e61290a4 | 1068 | mutex_init(&md->suspend_lock); |
2e93ccc1 | 1069 | spin_lock_init(&md->pushback_lock); |
1da177e4 LT |
1070 | rwlock_init(&md->map_lock); |
1071 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1072 | atomic_set(&md->open_count, 0); |
1da177e4 | 1073 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1074 | atomic_set(&md->uevent_seq, 0); |
1075 | INIT_LIST_HEAD(&md->uevent_list); | |
1076 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1077 | |
1078 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1079 | if (!md->queue) | |
6ed7ade8 | 1080 | goto bad_queue; |
1da177e4 LT |
1081 | |
1082 | md->queue->queuedata = md; | |
1083 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1084 | md->queue->backing_dev_info.congested_data = md; | |
1085 | blk_queue_make_request(md->queue, dm_request); | |
daef265f | 1086 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1087 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1088 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1089 | |
93d2341c | 1090 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1091 | if (!md->io_pool) |
6ed7ade8 | 1092 | goto bad_io_pool; |
1da177e4 | 1093 | |
93d2341c | 1094 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1095 | if (!md->tio_pool) |
6ed7ade8 | 1096 | goto bad_tio_pool; |
1da177e4 | 1097 | |
5972511b | 1098 | md->bs = bioset_create(16, 16); |
9faf400f SB |
1099 | if (!md->bs) |
1100 | goto bad_no_bioset; | |
1101 | ||
1da177e4 LT |
1102 | md->disk = alloc_disk(1); |
1103 | if (!md->disk) | |
6ed7ade8 | 1104 | goto bad_disk; |
1da177e4 | 1105 | |
f0b04115 JM |
1106 | atomic_set(&md->pending, 0); |
1107 | init_waitqueue_head(&md->wait); | |
1108 | init_waitqueue_head(&md->eventq); | |
1109 | ||
1da177e4 LT |
1110 | md->disk->major = _major; |
1111 | md->disk->first_minor = minor; | |
1112 | md->disk->fops = &dm_blk_dops; | |
1113 | md->disk->queue = md->queue; | |
1114 | md->disk->private_data = md; | |
1115 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1116 | add_disk(md->disk); | |
7e51f257 | 1117 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1118 | |
304f3f6a MB |
1119 | md->wq = create_singlethread_workqueue("kdmflush"); |
1120 | if (!md->wq) | |
1121 | goto bad_thread; | |
1122 | ||
ba61fdd1 | 1123 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1124 | spin_lock(&_minor_lock); |
ba61fdd1 | 1125 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1126 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1127 | |
1128 | BUG_ON(old_md != MINOR_ALLOCED); | |
1129 | ||
1da177e4 LT |
1130 | return md; |
1131 | ||
304f3f6a MB |
1132 | bad_thread: |
1133 | put_disk(md->disk); | |
6ed7ade8 | 1134 | bad_disk: |
9faf400f | 1135 | bioset_free(md->bs); |
6ed7ade8 | 1136 | bad_no_bioset: |
1da177e4 | 1137 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1138 | bad_tio_pool: |
1da177e4 | 1139 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1140 | bad_io_pool: |
1312f40e | 1141 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1142 | bad_queue: |
1da177e4 | 1143 | free_minor(minor); |
6ed7ade8 | 1144 | bad_minor: |
10da4f79 | 1145 | module_put(THIS_MODULE); |
6ed7ade8 | 1146 | bad_module_get: |
1da177e4 LT |
1147 | kfree(md); |
1148 | return NULL; | |
1149 | } | |
1150 | ||
ae9da83f JN |
1151 | static void unlock_fs(struct mapped_device *md); |
1152 | ||
1da177e4 LT |
1153 | static void free_dev(struct mapped_device *md) |
1154 | { | |
f331c029 | 1155 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1156 | |
d9dde59b | 1157 | if (md->suspended_bdev) { |
ae9da83f | 1158 | unlock_fs(md); |
d9dde59b JN |
1159 | bdput(md->suspended_bdev); |
1160 | } | |
304f3f6a | 1161 | destroy_workqueue(md->wq); |
1da177e4 LT |
1162 | mempool_destroy(md->tio_pool); |
1163 | mempool_destroy(md->io_pool); | |
9faf400f | 1164 | bioset_free(md->bs); |
1da177e4 | 1165 | del_gendisk(md->disk); |
63d94e48 | 1166 | free_minor(minor); |
fba9f90e JM |
1167 | |
1168 | spin_lock(&_minor_lock); | |
1169 | md->disk->private_data = NULL; | |
1170 | spin_unlock(&_minor_lock); | |
1171 | ||
1da177e4 | 1172 | put_disk(md->disk); |
1312f40e | 1173 | blk_cleanup_queue(md->queue); |
10da4f79 | 1174 | module_put(THIS_MODULE); |
1da177e4 LT |
1175 | kfree(md); |
1176 | } | |
1177 | ||
1178 | /* | |
1179 | * Bind a table to the device. | |
1180 | */ | |
1181 | static void event_callback(void *context) | |
1182 | { | |
7a8c3d3b MA |
1183 | unsigned long flags; |
1184 | LIST_HEAD(uevents); | |
1da177e4 LT |
1185 | struct mapped_device *md = (struct mapped_device *) context; |
1186 | ||
7a8c3d3b MA |
1187 | spin_lock_irqsave(&md->uevent_lock, flags); |
1188 | list_splice_init(&md->uevent_list, &uevents); | |
1189 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1190 | ||
ed9e1982 | 1191 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1192 | |
1da177e4 LT |
1193 | atomic_inc(&md->event_nr); |
1194 | wake_up(&md->eventq); | |
1195 | } | |
1196 | ||
4e90188b | 1197 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1198 | { |
4e90188b | 1199 | set_capacity(md->disk, size); |
1da177e4 | 1200 | |
1b1dcc1b | 1201 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1202 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1203 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1204 | } |
1205 | ||
1206 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1207 | { | |
165125e1 | 1208 | struct request_queue *q = md->queue; |
1da177e4 LT |
1209 | sector_t size; |
1210 | ||
1211 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1212 | |
1213 | /* | |
1214 | * Wipe any geometry if the size of the table changed. | |
1215 | */ | |
1216 | if (size != get_capacity(md->disk)) | |
1217 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1218 | ||
bfa152fa JN |
1219 | if (md->suspended_bdev) |
1220 | __set_size(md, size); | |
1da177e4 LT |
1221 | if (size == 0) |
1222 | return 0; | |
1223 | ||
2ca3310e AK |
1224 | dm_table_get(t); |
1225 | dm_table_event_callback(t, event_callback, md); | |
1226 | ||
1da177e4 LT |
1227 | write_lock(&md->map_lock); |
1228 | md->map = t; | |
2ca3310e | 1229 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1230 | write_unlock(&md->map_lock); |
1231 | ||
1da177e4 LT |
1232 | return 0; |
1233 | } | |
1234 | ||
1235 | static void __unbind(struct mapped_device *md) | |
1236 | { | |
1237 | struct dm_table *map = md->map; | |
1238 | ||
1239 | if (!map) | |
1240 | return; | |
1241 | ||
1242 | dm_table_event_callback(map, NULL, NULL); | |
1243 | write_lock(&md->map_lock); | |
1244 | md->map = NULL; | |
1245 | write_unlock(&md->map_lock); | |
1246 | dm_table_put(map); | |
1247 | } | |
1248 | ||
1249 | /* | |
1250 | * Constructor for a new device. | |
1251 | */ | |
2b06cfff | 1252 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1253 | { |
1254 | struct mapped_device *md; | |
1255 | ||
2b06cfff | 1256 | md = alloc_dev(minor); |
1da177e4 LT |
1257 | if (!md) |
1258 | return -ENXIO; | |
1259 | ||
1260 | *result = md; | |
1261 | return 0; | |
1262 | } | |
1263 | ||
637842cf | 1264 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1265 | { |
1266 | struct mapped_device *md; | |
1da177e4 LT |
1267 | unsigned minor = MINOR(dev); |
1268 | ||
1269 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1270 | return NULL; | |
1271 | ||
f32c10b0 | 1272 | spin_lock(&_minor_lock); |
1da177e4 LT |
1273 | |
1274 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1275 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1276 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1277 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1278 | md = NULL; |
fba9f90e JM |
1279 | goto out; |
1280 | } | |
1da177e4 | 1281 | |
fba9f90e | 1282 | out: |
f32c10b0 | 1283 | spin_unlock(&_minor_lock); |
1da177e4 | 1284 | |
637842cf DT |
1285 | return md; |
1286 | } | |
1287 | ||
d229a958 DT |
1288 | struct mapped_device *dm_get_md(dev_t dev) |
1289 | { | |
1290 | struct mapped_device *md = dm_find_md(dev); | |
1291 | ||
1292 | if (md) | |
1293 | dm_get(md); | |
1294 | ||
1295 | return md; | |
1296 | } | |
1297 | ||
9ade92a9 | 1298 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1299 | { |
9ade92a9 | 1300 | return md->interface_ptr; |
1da177e4 LT |
1301 | } |
1302 | ||
1303 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1304 | { | |
1305 | md->interface_ptr = ptr; | |
1306 | } | |
1307 | ||
1308 | void dm_get(struct mapped_device *md) | |
1309 | { | |
1310 | atomic_inc(&md->holders); | |
1311 | } | |
1312 | ||
72d94861 AK |
1313 | const char *dm_device_name(struct mapped_device *md) |
1314 | { | |
1315 | return md->name; | |
1316 | } | |
1317 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1318 | ||
1da177e4 LT |
1319 | void dm_put(struct mapped_device *md) |
1320 | { | |
1134e5ae | 1321 | struct dm_table *map; |
1da177e4 | 1322 | |
fba9f90e JM |
1323 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1324 | ||
f32c10b0 | 1325 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1326 | map = dm_get_table(md); |
f331c029 TH |
1327 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1328 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1329 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1330 | spin_unlock(&_minor_lock); |
cf222b37 | 1331 | if (!dm_suspended(md)) { |
1da177e4 LT |
1332 | dm_table_presuspend_targets(map); |
1333 | dm_table_postsuspend_targets(map); | |
1334 | } | |
1335 | __unbind(md); | |
1134e5ae | 1336 | dm_table_put(map); |
1da177e4 LT |
1337 | free_dev(md); |
1338 | } | |
1da177e4 | 1339 | } |
79eb885c | 1340 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1341 | |
46125c1c MB |
1342 | static int dm_wait_for_completion(struct mapped_device *md) |
1343 | { | |
1344 | int r = 0; | |
1345 | ||
1346 | while (1) { | |
1347 | set_current_state(TASK_INTERRUPTIBLE); | |
1348 | ||
1349 | smp_mb(); | |
1350 | if (!atomic_read(&md->pending)) | |
1351 | break; | |
1352 | ||
1353 | if (signal_pending(current)) { | |
1354 | r = -EINTR; | |
1355 | break; | |
1356 | } | |
1357 | ||
1358 | io_schedule(); | |
1359 | } | |
1360 | set_current_state(TASK_RUNNING); | |
1361 | ||
1362 | return r; | |
1363 | } | |
1364 | ||
1da177e4 LT |
1365 | /* |
1366 | * Process the deferred bios | |
1367 | */ | |
6d6f10df | 1368 | static void __flush_deferred_io(struct mapped_device *md) |
1da177e4 | 1369 | { |
6d6f10df | 1370 | struct bio *c; |
1da177e4 | 1371 | |
6d6f10df | 1372 | while ((c = bio_list_pop(&md->deferred))) { |
9e4e5f87 MB |
1373 | if (__split_bio(md, c)) |
1374 | bio_io_error(c); | |
1da177e4 | 1375 | } |
73d410c0 MB |
1376 | |
1377 | clear_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 LT |
1378 | } |
1379 | ||
6d6f10df MB |
1380 | static void __merge_pushback_list(struct mapped_device *md) |
1381 | { | |
1382 | unsigned long flags; | |
1383 | ||
1384 | spin_lock_irqsave(&md->pushback_lock, flags); | |
1385 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1386 | bio_list_merge_head(&md->deferred, &md->pushback); | |
1387 | bio_list_init(&md->pushback); | |
1388 | spin_unlock_irqrestore(&md->pushback_lock, flags); | |
1389 | } | |
1390 | ||
304f3f6a MB |
1391 | static void dm_wq_work(struct work_struct *work) |
1392 | { | |
1393 | struct dm_wq_req *req = container_of(work, struct dm_wq_req, work); | |
1394 | struct mapped_device *md = req->md; | |
1395 | ||
1396 | down_write(&md->io_lock); | |
1397 | switch (req->type) { | |
1398 | case DM_WQ_FLUSH_ALL: | |
1399 | __merge_pushback_list(md); | |
1400 | /* pass through */ | |
1401 | case DM_WQ_FLUSH_DEFERRED: | |
1402 | __flush_deferred_io(md); | |
1403 | break; | |
1404 | default: | |
1405 | DMERR("dm_wq_work: unrecognised work type %d", req->type); | |
1406 | BUG(); | |
1407 | } | |
1408 | up_write(&md->io_lock); | |
1409 | } | |
1410 | ||
1411 | static void dm_wq_queue(struct mapped_device *md, int type, void *context, | |
1412 | struct dm_wq_req *req) | |
1413 | { | |
1414 | req->type = type; | |
1415 | req->md = md; | |
1416 | req->context = context; | |
1417 | INIT_WORK(&req->work, dm_wq_work); | |
1418 | queue_work(md->wq, &req->work); | |
1419 | } | |
1420 | ||
1421 | static void dm_queue_flush(struct mapped_device *md, int type, void *context) | |
1422 | { | |
1423 | struct dm_wq_req req; | |
1424 | ||
1425 | dm_wq_queue(md, type, context, &req); | |
1426 | flush_workqueue(md->wq); | |
1427 | } | |
1428 | ||
1da177e4 LT |
1429 | /* |
1430 | * Swap in a new table (destroying old one). | |
1431 | */ | |
1432 | int dm_swap_table(struct mapped_device *md, struct dm_table *table) | |
1433 | { | |
93c534ae | 1434 | int r = -EINVAL; |
1da177e4 | 1435 | |
e61290a4 | 1436 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
1437 | |
1438 | /* device must be suspended */ | |
cf222b37 | 1439 | if (!dm_suspended(md)) |
93c534ae | 1440 | goto out; |
1da177e4 | 1441 | |
bfa152fa JN |
1442 | /* without bdev, the device size cannot be changed */ |
1443 | if (!md->suspended_bdev) | |
1444 | if (get_capacity(md->disk) != dm_table_get_size(table)) | |
1445 | goto out; | |
1446 | ||
1da177e4 LT |
1447 | __unbind(md); |
1448 | r = __bind(md, table); | |
1da177e4 | 1449 | |
93c534ae | 1450 | out: |
e61290a4 | 1451 | mutex_unlock(&md->suspend_lock); |
93c534ae | 1452 | return r; |
1da177e4 LT |
1453 | } |
1454 | ||
1455 | /* | |
1456 | * Functions to lock and unlock any filesystem running on the | |
1457 | * device. | |
1458 | */ | |
2ca3310e | 1459 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 1460 | { |
e39e2e95 | 1461 | int r; |
1da177e4 LT |
1462 | |
1463 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 1464 | |
e39e2e95 | 1465 | md->frozen_sb = freeze_bdev(md->suspended_bdev); |
dfbe03f6 | 1466 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 1467 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
1468 | md->frozen_sb = NULL; |
1469 | return r; | |
dfbe03f6 AK |
1470 | } |
1471 | ||
aa8d7c2f AK |
1472 | set_bit(DMF_FROZEN, &md->flags); |
1473 | ||
1da177e4 | 1474 | /* don't bdput right now, we don't want the bdev |
e39e2e95 | 1475 | * to go away while it is locked. |
1da177e4 LT |
1476 | */ |
1477 | return 0; | |
1478 | } | |
1479 | ||
2ca3310e | 1480 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 1481 | { |
aa8d7c2f AK |
1482 | if (!test_bit(DMF_FROZEN, &md->flags)) |
1483 | return; | |
1484 | ||
e39e2e95 | 1485 | thaw_bdev(md->suspended_bdev, md->frozen_sb); |
1da177e4 | 1486 | md->frozen_sb = NULL; |
aa8d7c2f | 1487 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
1488 | } |
1489 | ||
1490 | /* | |
1491 | * We need to be able to change a mapping table under a mounted | |
1492 | * filesystem. For example we might want to move some data in | |
1493 | * the background. Before the table can be swapped with | |
1494 | * dm_bind_table, dm_suspend must be called to flush any in | |
1495 | * flight bios and ensure that any further io gets deferred. | |
1496 | */ | |
a3d77d35 | 1497 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 1498 | { |
2ca3310e | 1499 | struct dm_table *map = NULL; |
1da177e4 | 1500 | DECLARE_WAITQUEUE(wait, current); |
46125c1c | 1501 | int r = 0; |
a3d77d35 | 1502 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 1503 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 1504 | |
e61290a4 | 1505 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1506 | |
73d410c0 MB |
1507 | if (dm_suspended(md)) { |
1508 | r = -EINVAL; | |
d287483d | 1509 | goto out_unlock; |
73d410c0 | 1510 | } |
1da177e4 LT |
1511 | |
1512 | map = dm_get_table(md); | |
1da177e4 | 1513 | |
2e93ccc1 KU |
1514 | /* |
1515 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
1516 | * This flag is cleared before dm_suspend returns. | |
1517 | */ | |
1518 | if (noflush) | |
1519 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1520 | ||
cf222b37 AK |
1521 | /* This does not get reverted if there's an error later. */ |
1522 | dm_table_presuspend_targets(map); | |
1523 | ||
bfa152fa JN |
1524 | /* bdget() can stall if the pending I/Os are not flushed */ |
1525 | if (!noflush) { | |
1526 | md->suspended_bdev = bdget_disk(md->disk, 0); | |
1527 | if (!md->suspended_bdev) { | |
1528 | DMWARN("bdget failed in dm_suspend"); | |
1529 | r = -ENOMEM; | |
1530 | goto flush_and_out; | |
1531 | } | |
e39e2e95 | 1532 | |
6d6f10df MB |
1533 | /* |
1534 | * Flush I/O to the device. noflush supersedes do_lockfs, | |
1535 | * because lock_fs() needs to flush I/Os. | |
1536 | */ | |
1537 | if (do_lockfs) { | |
1538 | r = lock_fs(md); | |
1539 | if (r) | |
1540 | goto out; | |
1541 | } | |
aa8d7c2f | 1542 | } |
1da177e4 LT |
1543 | |
1544 | /* | |
354e0071 | 1545 | * First we set the BLOCK_IO flag so no more ios will be mapped. |
1da177e4 | 1546 | */ |
2ca3310e AK |
1547 | down_write(&md->io_lock); |
1548 | set_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 | 1549 | |
1da177e4 | 1550 | add_wait_queue(&md->wait, &wait); |
2ca3310e | 1551 | up_write(&md->io_lock); |
1da177e4 LT |
1552 | |
1553 | /* unplug */ | |
2ca3310e | 1554 | if (map) |
1da177e4 | 1555 | dm_table_unplug_all(map); |
1da177e4 LT |
1556 | |
1557 | /* | |
46125c1c | 1558 | * Wait for the already-mapped ios to complete. |
1da177e4 | 1559 | */ |
46125c1c | 1560 | r = dm_wait_for_completion(md); |
1da177e4 | 1561 | |
2ca3310e | 1562 | down_write(&md->io_lock); |
1da177e4 LT |
1563 | remove_wait_queue(&md->wait, &wait); |
1564 | ||
6d6f10df MB |
1565 | if (noflush) |
1566 | __merge_pushback_list(md); | |
94d6351e | 1567 | up_write(&md->io_lock); |
2e93ccc1 | 1568 | |
1da177e4 | 1569 | /* were we interrupted ? */ |
46125c1c | 1570 | if (r < 0) { |
304f3f6a | 1571 | dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL); |
73d410c0 | 1572 | |
2ca3310e | 1573 | unlock_fs(md); |
2e93ccc1 | 1574 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1575 | } |
1da177e4 | 1576 | |
cf222b37 | 1577 | dm_table_postsuspend_targets(map); |
1da177e4 | 1578 | |
2ca3310e | 1579 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1580 | |
2e93ccc1 | 1581 | flush_and_out: |
304f3f6a | 1582 | if (r && noflush) |
2e93ccc1 KU |
1583 | /* |
1584 | * Because there may be already I/Os in the pushback list, | |
1585 | * flush them before return. | |
1586 | */ | |
304f3f6a | 1587 | dm_queue_flush(md, DM_WQ_FLUSH_ALL, NULL); |
2e93ccc1 | 1588 | |
2ca3310e | 1589 | out: |
e39e2e95 AK |
1590 | if (r && md->suspended_bdev) { |
1591 | bdput(md->suspended_bdev); | |
1592 | md->suspended_bdev = NULL; | |
1593 | } | |
1594 | ||
2ca3310e | 1595 | dm_table_put(map); |
d287483d AK |
1596 | |
1597 | out_unlock: | |
e61290a4 | 1598 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1599 | return r; |
1da177e4 LT |
1600 | } |
1601 | ||
1602 | int dm_resume(struct mapped_device *md) | |
1603 | { | |
cf222b37 | 1604 | int r = -EINVAL; |
cf222b37 | 1605 | struct dm_table *map = NULL; |
1da177e4 | 1606 | |
e61290a4 | 1607 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1608 | if (!dm_suspended(md)) |
cf222b37 | 1609 | goto out; |
cf222b37 AK |
1610 | |
1611 | map = dm_get_table(md); | |
2ca3310e | 1612 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1613 | goto out; |
1da177e4 | 1614 | |
8757b776 MB |
1615 | r = dm_table_resume_targets(map); |
1616 | if (r) | |
1617 | goto out; | |
2ca3310e | 1618 | |
304f3f6a | 1619 | dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL); |
2ca3310e AK |
1620 | |
1621 | unlock_fs(md); | |
1622 | ||
bfa152fa JN |
1623 | if (md->suspended_bdev) { |
1624 | bdput(md->suspended_bdev); | |
1625 | md->suspended_bdev = NULL; | |
1626 | } | |
e39e2e95 | 1627 | |
2ca3310e AK |
1628 | clear_bit(DMF_SUSPENDED, &md->flags); |
1629 | ||
1da177e4 | 1630 | dm_table_unplug_all(map); |
1da177e4 | 1631 | |
69267a30 | 1632 | dm_kobject_uevent(md); |
8560ed6f | 1633 | |
cf222b37 | 1634 | r = 0; |
2ca3310e | 1635 | |
cf222b37 AK |
1636 | out: |
1637 | dm_table_put(map); | |
e61290a4 | 1638 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1639 | |
cf222b37 | 1640 | return r; |
1da177e4 LT |
1641 | } |
1642 | ||
1643 | /*----------------------------------------------------------------- | |
1644 | * Event notification. | |
1645 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1646 | void dm_kobject_uevent(struct mapped_device *md) |
1647 | { | |
ed9e1982 | 1648 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1649 | } |
1650 | ||
7a8c3d3b MA |
1651 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1652 | { | |
1653 | return atomic_add_return(1, &md->uevent_seq); | |
1654 | } | |
1655 | ||
1da177e4 LT |
1656 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1657 | { | |
1658 | return atomic_read(&md->event_nr); | |
1659 | } | |
1660 | ||
1661 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1662 | { | |
1663 | return wait_event_interruptible(md->eventq, | |
1664 | (event_nr != atomic_read(&md->event_nr))); | |
1665 | } | |
1666 | ||
7a8c3d3b MA |
1667 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1668 | { | |
1669 | unsigned long flags; | |
1670 | ||
1671 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1672 | list_add(elist, &md->uevent_list); | |
1673 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1674 | } | |
1675 | ||
1da177e4 LT |
1676 | /* |
1677 | * The gendisk is only valid as long as you have a reference | |
1678 | * count on 'md'. | |
1679 | */ | |
1680 | struct gendisk *dm_disk(struct mapped_device *md) | |
1681 | { | |
1682 | return md->disk; | |
1683 | } | |
1684 | ||
1685 | int dm_suspended(struct mapped_device *md) | |
1686 | { | |
1687 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1688 | } | |
1689 | ||
2e93ccc1 KU |
1690 | int dm_noflush_suspending(struct dm_target *ti) |
1691 | { | |
1692 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1693 | int r = __noflush_suspending(md); | |
1694 | ||
1695 | dm_put(md); | |
1696 | ||
1697 | return r; | |
1698 | } | |
1699 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1700 | ||
1da177e4 LT |
1701 | static struct block_device_operations dm_blk_dops = { |
1702 | .open = dm_blk_open, | |
1703 | .release = dm_blk_close, | |
aa129a22 | 1704 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1705 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1706 | .owner = THIS_MODULE |
1707 | }; | |
1708 | ||
1709 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1710 | ||
1711 | /* | |
1712 | * module hooks | |
1713 | */ | |
1714 | module_init(dm_init); | |
1715 | module_exit(dm_exit); | |
1716 | ||
1717 | module_param(major, uint, 0); | |
1718 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1719 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1720 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1721 | MODULE_LICENSE("GPL"); |